46
Introduction As more and more companies acquire and store vast amounts of customer data, many of them have realized that this information may be useful in either generating additional revenue from their current customers (cross-sell and up-sell opportunities) or by using the data to better predict buying behavior that leads to future revenue streams. As a result, the field of data mining (or knowledge discovery) has spawned a vast array of new products, services, partnerships, and even new companies. In fact, because of the interest generated about this field of study, numerous established companies have jumped into the mix, resulting in an explosion of competition and interest. The result can be seen as both a blessing and a potential curse for any entrepreneur or company entering this business environment. On the one hand, money continues to be spent in this arena, even during the current decelerating economy. However, the competitive landscape has never been greater. Large, medium and small companies continue to battle one another for the corporate purse strings. In addition, as is often the case, the strong continue to get stronger, while weak (or financially strapped companies) either consolidate or vanish from the competitive landscape all together, making the prospects of surviving in this environment that much more sobering and challenging. Against this backdrop, the purpose of this report is to analyze and detail the marketing landscape of a proposed new predictive modeling technique. In addition to better understanding the current and future marketing climate, this report will also examine the numerous companies and competitors in the field, evaluating them for their strengths and weaknesses. Finally, this report will offer several possible strategic directions, along with the potential pitfalls associated with each choice. The ultimate purpose of the report is to determine the level of feasibility of the proposed predictive model. In other 1

Bearse Final

Embed Size (px)

Citation preview

Page 1: Bearse Final

IntroductionAs more and more companies acquire and store vast amounts of customer data, many of them have realized that this information may be useful in either generating additional revenue from their current customers (cross-sell and up-sell opportunities) or by using the data to better predict buying behavior that leads to future revenue streams. As a result, the field of data mining (or knowledge discovery) has spawned a vast array of new products, services, partnerships, and even new companies. In fact, because of the interest generated about this field of study, numerous established companies have jumped into the mix, resulting in an explosion of competition and interest.

The result can be seen as both a blessing and a potential curse for any entrepreneur or company entering this business environment. On the one hand, money continues to be spent in this arena, even during the current decelerating economy. However, the competitive landscape has never been greater. Large, medium and small companies continue to battle one another for the corporate purse strings. In addition, as is often the case, the strong continue to get stronger, while weak (or financially strapped companies) either consolidate or vanish from the competitive landscape all together, making the prospects of surviving in this environment that much more sobering and challenging.

Against this backdrop, the purpose of this report is to analyze and detail the marketing landscape of a proposed new predictive modeling technique. In addition to better understanding the current and future marketing climate, this report will also examine the numerous companies and competitors in the field, evaluating them for their strengths and weaknesses. Finally, this report will offer several possible strategic directions, along with the potential pitfalls associated with each choice.

The ultimate purpose of the report is to determine the level of feasibility of the proposed predictive model. In other words, can it successfully compete in the marketplace, and what are the challenges, risks and opportunities that might lie ahead?

1

Page 2: Bearse Final

Summary of AnalysisBelow are the key points affecting the feasibility of the proposed new predictive modeling technique:

1. Based on current research, the data mining market can be expected to grow in the immediate future and become an even more important business tool compared to current usage. As a result, there appears to be continued opportunity in the marketplace, especially for better and more accurate predictive methods.

2. Because of its potential value, a new and improved predictive model has wide applicability in all kinds of different markets, such as finance, manufacturing, telecommunications, health care, and transportation. As a result, there appear to be sizeable potential vertical markets for a new and improved predictive model.

3. Competition in the data mining market, however, is stiff. It includes not only numerous small-sized companies, but large, dominating competitors as well. In addition, non-statistical software companies have also entered the field. As a result, achieving useful brand awareness and sales penetration for a startup in the “retail” channel will likely be formidable and require significant financial resources.

4. Because of the highly competitive environment surrounding the larger software leaders in database management, data warehousing and CRM, such as Oracle and IBM, there appears to be an opportunity for these companies to bid up the price of securing the rights to new and improved predictive modeling methods.

5. In addition, because of the vast number of companies-large and small—using predictive models in data mining, there appears to be opportunity to license these methods at the “wholesale” level.

6. Primary marketing research is needed to fully and accurately ascertain the degree of interest and market potential of the proposed new modeling technique.

2

Page 3: Bearse Final

Marketing Climate AnalysisEconomic and Venture Capital ConditionsEconomic ConditionsWhile no one can accurately predict when the U.S. and world economies will pick up again, it does appear safe to say that the 3rd and 4th Quarters of this year will probably remain tepid. Dell Computer, for example, does not expect its business to pick up until the 1st Quarter of 2002.

If the U.S. and world economies continue to drift (or even falter further), it is likely to have an adverse impact on the sales and distribution of high-end data mining products. Companies continue to put off large software expenditures unless they are critical to the success of the organization. It could be argued that data mining software products may have an advantage over other products during a decelerating economy since they are intended to help companies save costs (by being more efficient in marketing efforts) or by increasing revenue (by finding more opportunities with current customers or by efficiently identifying new potential customers). But whether or not prospective customers can be motivated and convinced to purchase such software during a “down” economy remains to be seen. Current expenditures for all software products have slowed considerably.

Venture Capital ConditionsCurrent conditions to raise money in the venture capital market (8/01) remain difficult and are expected to continue. Not only has the market in general tightened up considerably after the fallout from the excessive investing spree in the past two years, but first round financing has become even tougher. A number of venture capital firms are either funding only second, third or fourth round offerings or not investing at all, preferring instead to “weather the storm” or support their current portfolio.

This is not to say that funding has dried up completely. However, venture capitalists can now be much more selective and careful, and are typically garnering better terms on their deals. Better due diligence has also returned to the venture capital community, which means greater scrutiny is occurring with respect to market potential, management capability, and exit strategies. This means that a new company built around a superior predictive model must likely show dramatic revenue and profit potential at a fairly early clip in order to attract the interest of venture capitalists today, and he impact on the current market must almost be seismic and revolutionary. The days of handing out large dollar amounts to shaky business plans are over.

Political and Regulatory ChangesNo political or regulatory changes now or in the “pipeline” appear to have any adverse impact on a new predictive modeling company. A continued drive, however, to make federal, state, county and city governments more consumer-centric (especially with the Internet) should aid in the continued development of better data mining and analytical products.

3

Page 4: Bearse Final

Industry and Market TrendsBelow is a list of possible industry trends that may have an impact on a new predictive modeling business.

Amount of data collected and stored by companies is likely to increase. Amount of data mined by companies is likely to increase. Data warehouse and database management software companies will continue to

“embed” analytics into their products. Desire for improved accuracy in data mining techniques is likely to continue. Growth in the business intelligence (BI) market is likely to continue. Growth in the CRM market is likely to continue. Trend towards greater personalization on Web will continue (privacy issues

aside). Integration of all customer “touch points” will continue. Integration of ERP and CRM offerings will continue. Data mining applications into e-business and Web will continue. Interest in real-time analytical capabilities will continue. Initiatives for predictive modeling standards (PMML) will continue. Integration of business intelligence (BI) offerings and data mining will continue. Integration of marketing campaign management offerings and data mining will

continue. Attention to customer retention and lifetime value (LTV) will continue. Continued move toward more open, not proprietary platforms will continue. Trend towards acceptance (albeit slow) of ASP model in data mining will

continue. Trend towards making data mining easier to use for non-statistician will continue. Consolidation in the CRM market likely will continue. Interest in reducing “churn” in telecommunications and other markets is likely to

continue.

Below is further elaboration on a select group of these industry and market trends.

Amount of data collected and stored by companies is likely to increase.Despite the current downturn in the economy, according to IDC, a global market intelligence and advisory firm, the data warehouse industry is still expected to grow from $5 billion in 1999 to $17 billion in 2004. The increase in data being collected and stored can, in part, be attributed to a greater number of “touch points,” where customers can make contact with companies at Web sites, through e-mail, and with call center agents via the telephone. As more data is collected and stored, more companies will likely continue to analyze that data to find some competitive advantage.

Amount of data mined by companies is likely to increase.More companies continue to increase their analytical capabilities in their software, including CRM, ERP, database management and data warehouse companies. In fact, IDC expects revenue in this market space to continue to grow, reaching the $6 billion figure in 2004 compared to today’s $2 billion figure. Fueling part of this trend is a need by marketers to improve the efficiency and

4

Page 5: Bearse Final

performance of their marketing efforts through pinpoint targeting. Thus, data mining is used to better identify prospective customers and match the product offering with their specific interests.

Data warehouse and database management software companies will continue to “embed” analytics into their products.This recent trend can be seen in the product configurations currently being offered by the big three database management companies—Oracle, IBM and Microsoft. Within the last two years all three companies have “embedded” their software with improved analytical capabilities, typically by partnering with a data mining company rather than building it themselves. While most of the analytics involve business intelligence capabilities, it is expected that the software companies will continue to improve the predictive modeling capacity of their offerings now and in the immediate future.

Desire for improved accuracy in data mining techniques is likely to continue.The capability of most data mining techniques of today still lack the precision that most marketers want and need in order to sell more while spending less. For example, most predictive models on the market today have inherent weaknesses in being able to make predictions outside of the historical data being analyzed. As a result, it’s safe to assume that companies within the predictive modeling space will continue to strive for improvement and efficiency in the products they use.

Integration of ERP and CRM offerings will continue.As a result of this integration between back-office and front-office functions, greater analytics will be required for companies to make better business decisions. In this instance, predictive modeling will be used not only in a marketing function, but it will be applied to other business intelligence functions as well, such as financial reporting and manufacturing.

Data mining applications into e-business and Web will continue. Interest in real-time analytical capabilities will continue.

Web analytics is clearly driving the desire to have real-time analytics, as business owners and marketers strive to satisfy customer interests instantaneously. Associated with this trend is the need for more effective analytical capabilities, which could conceivably be served by new and improved predictive modeling techniques.

ConclusionKey industry and market trends point favorably towards a new entrant in the predictive modeling environment. As more data is collected and stored and as competitive pressures continue to mount in all sectors of the marketplace, companies will seek to find competitive advantages by mining through their data in order to better predict and anticipate buying behavior. At the same time, continued usage of the Web will mean companies will also try to gain insight there as well in an effort to better target their prospective and current customers. Finally, the fact that more companies outside the “traditional” statistical software market (such as ERP, CRM and database management companies) continue to add analytical capabilities to their offerings suggests greater opportunities to license or sell predictive modeling software now and in the future.

5

Page 6: Bearse Final

There are risks inherent in such a conclusion. First, companies may be satisfied with the level of performance associated with the current predictive modeling offerings and not have much interest in improved modeling techniques. Second, in-house data mining and predictive modeling activities could prevent a new entrant from gaining access to prospective buyers at the wholesale level. Third, a backlash over privacy may derail efforts at further personalizing the Web, thereby decreasing the need for more data mining. And finally, continued growth in the data mining market likely means increased competition.

Overall, however, the gains appear to outweigh the risks, at least in terms of industry and market trends.

6

Page 7: Bearse Final

Competitive AnalysisThe competitive landscape for a proposed predicting modeling company can be best divided into two main groups:

1. Statistical software, data mining and analysis companies2. Companies adding or embedding data mining capabilities into their software –

such as ERP, CRM, database management, business intelligence, and business applications firms

Statistical Software, Data Mining and Analysis CompaniesThe statistical software, data mining and data analysis markets appear to be heavily populated with both large, medium and small companies (in terms of revenue). In fact, one could argue that the field is extremely crowded. For example, a search in Yahoo that is combined with a published list in DM Review, a magazine for the database marketing environment, produced more than 130 company names. While a number of these companies operate in small, specialized niche markets or are one-person firms, others are competing directly with one another in a fierce battle for dollars and customers.

NOTE: While there is considerable overlap among statistical software, data mining and data analysis companies, statistical software companies are defined for this report as those companies whose principal business initiated with statistical packages and who later moved into other areas, such as data mining.

Statistical SoftwareTwo main players dominate this space – SAS and SPSS. While one is a public corporation (SPSS) and other private (SAS), both have considerable resources at their disposal, and thus are considered major competitors to any new company intending to directly compete for customers. A detailed summary of each company is outlined below.

Three other companies currently compete directly in the statistical software industry – Minitab, StatSoft and MathSoft (now called Insightful Corp.).

Originally developed in 1972 to help professors teach basic statistics, Minitab has grown to provide statistical packages to the corporate world as well. Its list of corporate customers include Ford Motor Company, 3M, Honeywell International, and General Motors. Based on its web site (www.minitab.com), the company appears to have had some success in the manufacturing industry, especially in concert with quality improvement. Its product was also named the winner in the 2001 Reader’s Choice Awards Statistical Software category by Science Computing and Instrumentation, a publication covering computing and analytical instrument technology.

The price for its main product (Minitab 13) sells for $995.

StatSoft, Inc. is a closely held corporation with headquarters in Tulsa, Oklahoma, and also grew out of a partnership of university professors and scientists. While the company claims the exact number of users of StatSoft products is difficult to determine, the

7

Page 8: Bearse Final

company estimates users of over 600,000 worldwide (universities/research institutions: 30%; corporations/manufacturing facilities: 60%; government agencies: 10%. Its primary statistical product is STATISTICA. On its website (www.statsoft.com), the company has a long list of corporate clients, including: Fastenal Company, Federal Express, FAA (Federal Aviation Administration), Fina Oil and Chemical, Ford Motor Company, Franklin Electric, Franklin Mint Trading Corporation, Friskie, Frito-Lay, Inc., Fuji, GAF Materials Corporation, Gap, Inc., GAST Manufacturing Corporation, General Mills Inc., General Motors, Georgia Pacific, and the Gillette Company.

STATISTICA Base sells for $795.

Insightful Corp. emerged from the newly divested Engineering and Education Products Division of MathSoft. Focusing principally on solutions for predictive analysis, data mining and business intelligence, Insightful generated $15.2 million in revenue in 2000 with its principle flagship product – S-PLUS. On its website (www.insightful.com), the company lists areas where its products have been established in the fields of finance, e-business, biopharm, manufacturing, GIS, and academia.

SASSAS calls itself the world’s leader in business intelligence software and services and who can argue? Founded in 1976, SAS serves more than 37,000 business, government and university sites in 111 countries and, according to its website (www.sas.com) 98 percent of the top 100 companies on the Fortune 500 are SAS customers. It has also experienced 24 years of double-digit growth with more than $1.2 billion in revenue, making it the largest private software company in the world.

The company’s product line (which has more than 40 distinct products) includes applications in data warehousing, data mining, online analytical processing (OLAP), collaborative technologies, integration technologies, and CRM.

The largest percentage of SAS’s revenue came from the financial and insurance industries (25%). Another 14% of its revenue comes from government clients, while 14% also comes from manufacturing companies. Telecom companies account for 12% of SAS’s revenues, while sales and service, health care and education comprise the rest of its revenue stream.

SAS’s growth in 2000 appeared to result from a strong showing in e-business, customer relationship (CRM) and supplier relationship management (SRM). CRM, SRM and enterprise performance management accounted for 75% of the company’s new business in 2000. Two areas the company appears to be focusing on for future growth are the delivery of near real-time data and results and marketing automation, which would be intimately tied to the data mining and analysis products.

SAS has also begun to develop greater alliances with large hardware and software computer firms, such as IBM. In August, both companies announced a joint marketing alliance, which will combine SAS’s data mining software with IBM’s computer hardware

8

Page 9: Bearse Final

and services. It also has recently begun to purchase smaller companies that had complementary products and services, such as DataFlux and Intrinsic.

From a sales perspective, SAS is a formidable competitor, with a rapidly expanding direct sales force worldwide. In 2001, for example, the company plans to increase its sales force in the Americas by 150 percent and add more than 1,000 employees worldwide.

The company’s flagship product for data mining and predictive modeling is Enterprise Miner, a data mining solution that has a full range of integrated predictive models and algorithms – including decision trees, neural networks, regression, memory based reasoning, bagging and boosting ensembles, two-stage models, clustering, time series, and associations. This modeling “cornucopia” forms in large part the basis for the company’ positioning strategy.

SAS’s pricing for software ranges from $2,500 a year for personal computer usage to six figures for a mainframe, according to the company’s chief marketing officer. The company relies heavily on a yearly “subscription fee” from its clients, thereby generating a steady rate of revenue month after month.

According to published reports, the company plows 30 percent of its revenue ($30 million) each year into research and development.

SAS must be considered a considerable competitor in the field of data mining, predictive modeling, and data warehousing. In fact, it could be considered the “800-pound gorilla” in the marketplace because of its sheer size, scope and success. Because they are a privately-held company, SAS is much more able to respond to the marketplace in relative secrecy or develop new products through R&D. It has also been spectacularly successful in marketing to large, deep-pocketed companies, especially financial institutions and enjoys the fruit of a world-wide sales force.

However, it is not without its weaknesses. Much of its data mining software is built from within, meaning that its insularity may slow true progress and innovation. In addition, this dependence on internal development may also prevent it from acquiring new and better predictive modeling. Second, its size now means it might be less nimble to respond to the marketplace than in previous years. Third, its proprietary systems make it vulnerable to open standard trends. Fourth, because the company is so heavily dependent on a few “golden” markets, it may be unwilling or unable to develop other potentially beneficial markets. Fifth, its predictive modeling techniques may currently not be considered “state-of-the-art” among modeling experts.

Conclusion: Competing directly against SAS in its own “backyard” may not be the most prudent business decision. However, a new company might be able to compete directly against this company on the overwhelming evidence of superiority in its predictive modeling techniques. It might also be able to compete by adding

9

Page 10: Bearse Final

significant customer value with a slightly lower-priced product, thus providing greater value through those two elements (better product + better price) and matching SAS’s legendary customer service. It’s highly doubtful, however, that SAS’s current customers (which includes practically every Fortune 100 company) would switch providers, especially after having invested enormous amounts of time and energy in working with SAS and after receiving exemplary customer care. If a new predictive modeling company wanted to avoid competing against SAS, it is suggested it find a niche small enough or “stealthy” enough to not be on SAS’s current radar screen.

SAS also does not appear to be a good target for a buyout, if only because of the “not-invented-here” syndrome. In other words, if all of its predictive modeling capabilities are created “in-house,” then there may be less motivation to purchase (either outright or through licensing) another predictive modeling technique. While it has recently begun to purchase other companies, those purchases complemented SAS’s core products and did not directly compete. Thus, there is no purchase history of a potentially competitive company.

SPSSSPSS is a publicly traded company located in Chicago, Illinois. Started by three Stanford graduate students, SPSS has grown to be a 900-person company with more than $140 million in annual revenue. According to the company’s annual report (10K) in April, 2001, approximately two-thirds of the Company’s customers are commercial organizations. Among its customers in the public sector, SPSS’s offerings are primarily used to improve interactions between government agencies and their constituents, as well as detect fraud and other forms of non-compliance.

The company’s main offerings include: CustomerCentric, an analytical solution specifically for CRM applications; Clementine, a general data mining and analysis product; and the MR Dimensions and Quantime product line for use by the professional market research market.

SPSS markets and sells its solutions, products and services primarily through worldwide field sales and telesales organizations. Historically, product sales have been made by the telesales organizations from leads driven by advertising, direct mail, tradeshow attendance, and customer references. The company has 60 licensed distributors. Student versions of its software are published by Prentice Hall and its sales representatives working directly with faculty on college campuses worldwide.

Net revenues increased from $123 million in 1998 to $141 million in 1999, while dropping to $140 million due to a change in accounting practices. Revenues dropped 25% for the three months ending March 31 to $36.5 million. Net loss totaled $10.5 million during that same period, which includes the inclusion of a $7.8 million merger expense. Sales and marketing expenses for that same period (1998-2000) increased from $60 million to $81 million, and increase of 32%. The company attributed increases to expanding sales management and representatives.

10

Page 11: Bearse Final

According to the management of SPSS, vendors providing spreadsheets, query and reporting tools, and OLAP capability serve the largest part of the general market for data analysis software. In addition, the company has increased its emphasis on the sale of higher-priced products and analytical solutions.

For the SPSS product line, list prices for single-user licenses of desktop products are about $999. List prices for annual licenses for products on mainframes, minicomputers, UNIX workstations and Windows NT servers range from $4,500 to $15,000, while perpetual licenses run from $10,000 to over $30,000. Clementine typically sells between $50,000 and $75,000. CustomerCentric usually costs between $750,000 and $1.5 million.

Like SAS, SPSS has announced a number of strategic marketing partnerships. In July, the company announced that it joined the Siebel Alliance Program as a strategic software partner. SPSS’s analytical solutions and products are expected to be integrated with Siebel eBusiness Applications to support enhanced customer segmentation and target marketing. SPSS has also announced partnerships with Lucent, Unisys, and Azerity, a sells-die business software solutions provider.

On its web page (www.spss.com), the company lists the following vertical markets as areas of involvement: banking, e-business, government, health care, higher education, manufacturing, market research, primary and secondary education, retail, telecommunications and transportation.

SPSS is another strong competitor, especially in the smaller statistical software market. It has proven to be able to grow its business from a university-focused market to a corporate strategy that now includes e-business and higher-end data mining opportunities. Although the company is likely to face considerable competitive pressures within the high-end market (principally from SAS), SPSS’s reputation for solid products and fair prices make it a major player in whatever market it goes after. SPSS also appears to have an extremely strong brand awareness, especially among statisticians, professors, and marketing research professionals. It is almost a de-facto standard in certain smaller markets, such as universities.

On the other hand, SPSS does not appear to be taking a leadership position when it comes to the most advanced data mining and predictive modeling techniques. Its offerings appear to be fairly standard models and applications, and the company appears content to maintain its status quo, despite a “healthy” R&D budget. As a result, if a new company wanted to compete directly against SPSS, a case could be made for competing on superiority, much like SAS. But that same company would be under enormous financial pressures to match SPSS’s sales, marketing and distribution capabilities which cost on average $80 million a year. Only significant “buzz” (word-of-mouth) could likely counter such competitive strength and the success rate of such efforts are typically low.

11

Page 12: Bearse Final

Data Mining and AnalysisGenetic AlgorithmsOf the hundreds of companies listed on the Web for data mining, four appear to be using genetic algorithms as a differentiating characteristic for their positioning efforts – MineTech, Ward Systems Group, NeuralWare and NuTech Solutions . Others may be using genetic algorithms and using this technique as a clear differentiator, but they have not yet been identified.

MineTech is a private, New York-based startup company selling GMAX, a genetic algorithm-based predictive modeling technique that the company claims is faster, more flexible, and more accurate than current modeling techniques. The company was founded in 1999 by Kenn Devane, a 15-year veteran of the database marketing industry.

Because so little information has been written about the company or because of the information available on its web site is short on detail, little more can be said about its customers, direction, markets or positioning. It can be concluded, however, that the company’s overall awareness within data mining circles appears to be still fairly limited.

Prices are not listed on the company web site (www.minetech.com).

Ward Systems Group of Frederick, Maryland, boasts on its web site (www.wardsystems.com) that it is “the leader in artificial intelligence.” The private company has a number of different products available for the data mining market. The primary products are: NeuroShell Predictor, NeuroShell Classifier, Gene Hunter, and AI Trilogy, which combines the first three products into a single package. Gene Hunter uses genetic algorithms as the basis for its data mining capability.

Companies listed on Ward’s web site as customers include:American Bank and Trust, Chevron, Citigroup, Coca Cola, Dell, EPA, FBI, Coors, Duke Power, Digital Equipment, Frito-Lay, Glidden, GTE and Microsoft.

NeuroShell Predictor and NeuroShell Classifier are priced at $395. Gene Hunter is priced at $295. AI Trilogy is priced at $995.

NeuralWare (www.neuralware.com) is Pittsburgh-based company specializing in neural networks to solve data mining problems. The company’s flagship product is NeuralWare Predict (formerly NeuralSIM). Used as an Add-in to Microsoft’s Excel spreadsheet product, Predict uses genetic algorithms to build and evaluate “mini-networks to identify not only which domain inputs are significant, but also the type of transform function that ultimately produces the best network.”

Predict sells for between $1,995 and $4,995, depending on the processor used in the computations.

NuTech Solutions is a Charlotte-based company started by a computer science professor and his son and it specializes in solutions for data acquisition, data mining, system

12

Page 13: Bearse Final

optimization, knowledge management, and network infrastructure. The company’s web site (www.nutechsolutions.com) indicates that the different types of technologies used include Genetic algorithms, neural networks, and fuzzy systems. Its primary business focus is on risk management and predictive analysis, supply chain optimization, knowledge management, tools, and custom solutions.

The client list as indicated on the company web site is large, impressive and global. Among the companies listed are: BASF Corporation, BMW Group, Daimler Chrysler, Ford, Federal Express, Carolina Power and Electric, Wal-Mart, NASA, and the Polish National Air Force.

The company web site also indicates how its technology is used by its customers. Some of the uses include: development of an optimization tool, optimization of airfoil designs for military aircraft, development of an expert system for auditing tax returns, crew scheduling optimization, data mining, and optimization of fuel assembly designs for nuclear reactors.

Prices for the products delivered or services rendered y NuTech Solutions are not listed on its web site. However, it appears the work is customized for each customer and thus likely to have a high price tag.

Although currently a private company, revenues are estimated to be in the $20-30 million range annually.

Other Data Mining CompaniesThe number of companies that focus almost exclusively on data mining is large and diverse. For example, a recent entrant, digiMine, focuses primarily on e-commerce companies and provides outsourced data mining capability. On the other hand, Cognos offers “business intelligence” for CRM environments, while Axonal Health Solutions (recently acquired by CareSteps) provides predictive modeling and data mining capabilities primarily for the health care industry.

While the number of companies in the data mining field are too numerous to mention in this report (see appendix), there are several that appear to be getting higher press coverage than others. They are: Cygron, Cognos, Angoss, Magnify, HNC, Mantas, Pilot Software, Sagent, Salford Systems, Quadstone, Unica, Accrue, Inxight, Evoke Software, digiMine, and WhiteCross Systems.

Embedded CompaniesData Warehouse and Database Management CompaniesSome of the most ambitious uses of data mining have occurred with data warehouse and database management companies. Not only have they incorporated commonly-used data mining software into their hardware or software offerings, but they have begun to aggressively promote these attributes in their promotional literature.

13

Page 14: Bearse Final

Data Warehousing CompaniesThe largest of the data warehousing companies all have embedded data mining and analytic capabilities into their offerings. Chief among these companies are IBM, Microstrategy, NCR, Oracle, Platinum and SAS.

Database Management CompaniesIn terms of embedding their current offerings, the same can be said for database management companies as was said about data warehouse companies. All of the top companies have now imbedded their software with analytic capability. These companies include: IBM, Microsoft, Oracle, and Sybase. It also includes Acxiom, Broadvase, Informatica, Microstrategy, SAS and Sybase as well.

Enterprise Resource Planning (ERP) and Supply Chain CompaniesThose companies adding analytical and data mining capabilities to their ERP and supply chain technology offerings include: SAP, Oracle, Baan, Manugistics, J.D. Edwards, Ariba, Commerce One, Ariba, and i2 Technologies.

Business Intelligence (BI) CompaniesBusiness intelligence is an analytical capability focused on primary business functions, such as sales and manufacturing reporting and query. Those companies specializing in BI have increased their software’s capability in recent years by adding data mining functions. The companies included in this category include: Microsoft, Oracle, Hyperion, Microstrategy, Business Objects, Cognos, SAS, SAP, Sybase, Spotfire, Informatic, Crystal Decisions, nQuire, and White Light.

Customer Relationship Management (CRM) CompaniesCustomer relationship management (CRM) refers principally to those companies using technology to help their clients acquire, sell to, and retain customers through multiple touch points in the hopes of building long-lasting relationships. Companies specializing in sales, marketing and customer service automation are now embedding their software with data mining capabilities. Some of those companies include: BroadVision, Clarify, HNC Software, Oracle, PeopleSoft, Onyx Software, Siebel systems, ServiceSoft, eGain, eShare, Unica, Silknet, Accrue, Kanna (and Broadbase), E.piphany, and Attune.

Potential SuitorsWhile it’s premature to consider suitors for a new predictive modeling company before anything is built or marketed, it may be prudent to keep several companies on the radar screen early.

Certainly, any software or hardware company with deep pockets, large customer bases, and aggressive marketing strategies must be considered a potential candidate. Thus, companies like IBM, Oracle, Microsoft, NCR, SAP, Siebel and SAS immediately come to mind.

14

Page 15: Bearse Final

PeopleSoft should probably also be considered since it not only has successfully integrated its purchase of Vantive’s CRM products into its own software, but its recent financial results indicate a strong turnaround and skillful management.

SPSS should also be considered a possible suitor, if for no other reason than their desire to move into the high-end data mining market to compete with SAS. A superior predictive modeling product may be just what the company needs to effectively compete.

Of the pure Internet companies, e.piphany appears to continue to build market share and might be a potential suitor.

15

Page 16: Bearse Final

MarketsThe potential markets for data mining software appears large and diverse. Because so many companies are trying to make sense out of their collected data, data mining efforts cover practically every industry.

There are, however, several vertical markets that appear regularly in most of the data mining literature. They are:Financial (banks/credit card)Financial (insurance)Telecommunications Investments and SecuritiesTravel and TourismTechnologyEntertainmentGovernment Manufacturing TransportationUtilitiesRetailHealth CarePharmaceuticalBiotechComputers and TechnologyAerospaceE-businessAgricultureEducationMarket ResearchDatabase Marketing and Catalogue SalesCRM (Customer Relationship Management)Scientific Research

Further explanation of selected markets:

Financial (banks/credit card)Banks are probably the largest user of data mining techniques, specifically to better pinpoint target marketing efforts related to the vast credit card industry. For example, financial institutions generate 25 percent of SAS’s annual revenues, which is far ahead of any other category. Also, in a poll of those in the data mining industry, 17% said they had applied data mining techniques to banking, compared to 15 percent for e-commerce and 11 percent for telecommunications issues. (Source: KD Nuggets)

The financial industry uses data mining and predictive modeling techniques to attract, acquire new customers and to retain and improve profitability with current customers. Specifically, predictive modeling is used by banks as part of its direct mail campaign to

16

Page 17: Bearse Final

attract new credit card users. Based on profiles of current customers, along with demographic and psychographic data, the bank can send targeted credit card offers to those people who it believes are more likely to respond. And because the credit card industry represents a significant source of revenue for banks, data mining and predictive modeling have become important components of a bank’s database marketing efforts.

Data mining is also used within the banking industry to cross-sell and up-sell current customers and to try to predict which customers are likely to stop using the bank’s services. In addition, banks segment customers and create tailored strategies based on credit risk, behavior, lifestyle, utilization and channel preference. By harnessing the patterns that emerge from its massive database of customers, banks have been able to increase revenues and profits with more targeted marketing efforts.

The typical buyer for data mining products within the financial industry is a vice president of database marketing or senior statistician. While both positions come at a data mining purchase from different perspectives, both have the same need—to optimize its database marketing efforts with a product of great value and utility. The products are sold typically with a long sales cycle (six months or longer) and a direct selling method is commonly used.

Financial (insurance)Data mining is used by insurance companies to solve a number of problems. Chief among them is establishing rates, acquiring new customers, retaining current customers, and detecting fraudulent claims. Modern data mining models can accurately predict risk, therefore allowing insurance companies to set rates more accurately, which in turn results in lower costs and greater profits. Also, by using data mining methods the insurance industry can identifying in segments those customers with the highest life-time value, and then targeting prospective customers who have characteristics similar to those most valued customers. Predictive modeling can also be used to identify those customers who are most likely to be candidates for switching to another carrier, and thus can be contacted prior to the expiration date of their policy.

TelecommunicationsThe key need for using data mining and predictive modeling techniques in the telecommunications industry is churn. Churn refers to the turnover of customers from one provider to another. Estimates vary as to the full extent of churn in the telecom industry, but they typically range from 20-40 percent or more. If companies can reduce their volume of churn, revenues are increased and costs are lowered as acquisition costs are amortized over a number of years.

Data mining and predictive modeling are used to determine which people will leave, when they’ll leave and why. In addition, companies use those techniques to determine which customers are the best to keep and which ones may offer the best lifetime value.

17

Page 18: Bearse Final

GovernmentGovernmental agencies are using data mining and predictive modeling techniques principally for fraud detection. With significant dollar amounts being distributed through governmental programs, such as Medicare and Medicaid, the U.S. government continues to seek ways to reduce the amount of fraud being perpetrated on the system. Along the same lines, the Internal Revenue Service is using data mining and predictive modeling to develop a more precise manner in which to conduct audits.

CRMAs one of the hottest topics in business and technology circles in the past several years, the customer relationship management (CRM) market has proven to be a boon to data mining. Companies specializing in CRM (which is a business strategy to develop and maintain long-term relationships with customers) are combining sales force, marketing management, and call center automation with data analytics in an attempt to provide one-on-one, personalized service to customers, regardless of the size of the customer base. Data mining and predictive modeling are used to anticipate the needs of customers by analyzing current and past buying behavior or patterns. The market for CRM products and services is expected to continue, as is the need for data mining. IDC estimates the CRM applications market to total $6.2 billion in 2000 and grow to $14 billion by 2005.

E-businessThe proliferation of web sites, along with the significant data flow associated with those sites, has spawned the growth of web tracking and analysis. Data mining and predictive modeling techniques are being used by companies to determine (a) who is coming to a particular site and (b) how to better target those prospective customers with personalized messages and offerings that appeal just to them. Although privacy concerns continue to unfavorably bring these efforts into the public spotlight, these efforts continue unabated and are likely to become more important as companies strive to make their marketing efforts more efficient and effective. Several data mining companies have specialized in this market for several years now, while others have only recently entered the picture, such as SAS.

Another area on the web that will require the use of improved data mining techniques is in near real-time analysis—a trend that continues to gather steam. Companies are anxious to get instant feedback on their marketing efforts so they can quickly change or adopt new strategies.

18

Page 19: Bearse Final

Pricing AnalysisPricing in the data mining and predictive model markets has a tremendous range. On the low end, companies such as Ward Systems Group and SPSS are offering statistical software packages that cost under $1,000. At the opposite end, companies such as SAS are offering their software packages for fees as high as $100,000 a year or more. Meanwhile, a number of companies are operating with a pricing strategy somewhere in the middle.

19

Page 20: Bearse Final

Marketing StrategyMarketing ChoicesAny new company has several strategic marketing choices to make.

1. Distribution One of the first key marketing decisions that must be made is whether or not a new predictive modeling solution would be sold or licensed at the “retail” or “wholesale” level. In other words, would the solution be sold or licensed directly to the users—such as financial institutions and manufacturers—or would it be sold or licensed to other companies in the data mining field who are currently providing services to those users and are looking for better solutions?

“Retail”The primary advantages of distributing at the “retail” level are several.

Retail distribution allows the proposed company to satisfy direct needs in the marketplace, especially if it can demonstrate a superior product, and have direct access to customers and their feedback.

Selling directly to users allows the proposed company to shift resources and direction, depending on which market appears to be strongest and most profitable.

A direct selling approach provides higher visibility the marketplace, thereby allowing the company to capitalize on strong word-of-mouth and promotional opportunities.

The disadvantages of selling directly involve several key areas. A direct approach requires a significant selling infrastructure nationally and

internationally, including extensive costs for experienced sales representatives and back-end support.

Selling at the retail level directly cuts into the teeth of competition. Some competitors are well-financed and already heavily entrenched in certain markets.

A “retail” approach creates press attention, some of which may be negative (i.e. product reviews).

The “retail space in general is already heavily populated with data mining companies who are all fighting for the same corporate customers.

Greater resources are likely needed to generate a sufficient level of brand awareness, such as the use of advertising, promotions, and public relations.

“Wholesale”The possible advantages to selling to the providers of data mining services are several.

A much smaller sales force would likely be necessary since the total number of potential buyers is smaller than at the retail level.

20

Page 21: Bearse Final

There’s likely to be less competition at the wholesale level. In fact, it might even be suggested that there are few competitors offering predictive modeling products and services to data mining providers.

The wholesale level might have a faster sales cycle. Data mining firms may be inclined to purchase a model that quickly gave them a competitive advantage.

There’s greater opportunity for exclusivity, resulting in a higher-priced sale. Competitive forces within the data mining community may bid up the price.

Fewer marketing resources are likely to be needed to generate awareness compared to retail distribution.

There are also several disadvantages associated with selling new predictive modeling techniques at the “wholesale” level. Data mining companies may only consider in-house predictive modeling

work and not use models from an outside source. The total market size may be narrowed considerably compared to a “retail”

environment. Selling to a “middleman” would likely mean reduced or “wholesale” prices. Greater scrutiny or proof of concept may occur, especially if the sale is to

statisticians or experienced data miners.

2. Target Markets A second decision relates to which markets to attack. If a retail direction is taken, would the new company go after more highly focused market, such as manufacturing, or should it take a more general approach and make its products and services available to any and all potential customers? At the wholesale level, does the proposed company market its services to all data mining companies, or does it restrict availability to only those large and deep-pocketed players who can afford a conceivably large price tag for a new and improved predictive modeling technique? Furthermore, should Web-centric companies be the target of the marketing efforts for the new model? Which target markets are likely to generate an optimal level of revenue and profitability?

Financial and Governmental While deep-pocketed and a big user of data mining software, the financial market also appears to be extremely competitive and dominated by SAS. Thus, unless one has a clear competitive advantage, can successfully position itself as being better and different in communication strategies, and is willing or able to apply the necessary marketing resources to the effort, one might want to steer clear of this market. The same might be said for the government market, where SAS also appears to have a significant share of the market. However, because of a more rigid RFP process, it is possible that a startup company could make inroads against its larger competitors, especially if it can compete on lower price and superior product.

Telecommunications

21

Page 22: Bearse Final

Similar to the financial market, the telecommunications industry is highly competitive and attractive to large competitors in the data mining field, such as SAS. At the same time, the need to reduce churn and maximize a customer’s lifetime value (LTV) has never been greater. As more services and benefits are added wireless offerings in the US. and abroad, a lot of money is at stake. Thus, any new predictive model which can more accurately predict potential dropouts from the service or identify better customers is likely to be met with great enthusiasm in the marketplace.

ManufacturingLarge manufacturing concerns may hold significant promise for a startup, especially considering that another data mining company (NuTech Solutions) appears to have demonstrated the viability of this approach. If manufacturers can use new predictive modeling techniques to “optimize” their operations and distribution efforts, significant dollars can be saved. These large cost savings, in turn, justify the expense for the predictive modeling software. While certainly highly competitive and smaller than the services industry, manufacturing appears to offer significant opportunity.

Pharmaceutical, Biotech and Health CareThe same might be said for the pharmaceutical and biotechnology markets, which rely heavily on sophisticated quantitative analysis to create new drugs. Any model that could improve the success rate for both the development and distribution of pharmaceuticals worldwide would likely be seen as a valuable product or service and thus marketable. This market, however, is also currently hotly competitive, with large data mining companies, such as SAS, generating significant revenue already.

CRMWhile growing in popularity, CRM (customer relationship management) as a market for a new predictive model may be a blessing and a curse. It’s a blessing in that more and more companies are clamoring for better and more effective predictive models. And since money continues to be spent in this market, that means there is money to be made by aggressive, successful businesses. However, hype continues to surround this market. As a result, unrealistic expectations in the marketplace may set up any entrant for possible failure, unless there is a clear, demonstrable breakthrough. In addition, this space has become crowded with large and small players, suggesting a continuation of the consolidation that began in 2000 and continues today. A new company would need a significant brand awareness campaign to rise above the “noise.”

EducationThe education market seems heavily saturated with well-entrenched players, such as SPSS and StatSoft. In addition, pricing remains low, which in turn keeps overall revenue totals much lower. Any new entrant would probably be wise to avoid this market, unless there were legitimate “credibility” reasons to do so.

22

Page 23: Bearse Final

Database Marketing (Direct Mail)The database marketing also appears to be heavily crowded with competitors, including giants such as SAS. However, it is a significant business opportunity since companies spend billions and billions of dollars each year to distribute direct mail literature, including catalogues. A new entrant in the data mining field could probably successfully compete in this arena if it positions itself as being able to dramatically improve the “lift” of the database marketing efforts as a whole.

Market Research and Scientific ResearchThe market research market does not appear to be a significant opportunity for a new entrant. SPSS has a large share of this market and appears to be satisfying the needs of market researchers with its current products. In addition, the revenue generated per sale is relatively small, and the overall market size is also relatively small (with about 5,000 market research firms and individual practitioners nationwide). The same could be said for scientific research as well, although the stature of the product may be raised with its use in the scientific community.

AgricultureAgriculture as a market may hold some promise for a new entrant. That’s because it may fall under the radar of other more well-established data mining firms and thus less competitive. Agriculture is also big business, especially worldwide, and highly dependent on accurate forecasting.

Travel, Tourism and EntertainmentTravel, tourism and entertainment may offer significant opportunity in the marketplace for a new predictive model entrant for a couple of reasons. First, database marketing is used fairly extensively by some of the larger tourism properties, such as cruise lines. Thus, any improvement in their targeted marketing efforts would likely be welcomed. Second, many of the top players in the field—such as SAS and SPSS—do not appear to be as well entrenched as in other markets, such as finance and telecommunications. Thus, establishing a beachhead in these markets may be a good strategy for a new entrant. It must be noted, however, that despite enormous dollars spent on entertainment in this country and abroad, the business as a whole does not generally embrace a more scientific and quantitative analysis approach to solving its business problems, as evidenced by the movie industry’s continued use of “gut instinct” and “limited focus groups” to pick movies.

E-BusinessE-Business offers some promise as a market for a new predictive model. However, in addition to being currently overcrowded with large and small competitors, the industry continues to go through the aches and pains of “growing up.” That means the financial viability of this market as a whole is still largely undetermined. There also continues to be a significant amount of hype associated with the market, and concerns over privacy may dampen any enthusiasm for improved data mining techniques. All things being equal, a new entrant may want to avoid putting large

23

Page 24: Bearse Final

resources into this market, unless high brand awareness and a demonstrable edge generates into large market share in a relatively short period of time.

3. Pricing A third decision relates to pricing. In general, where should the proposed product be priced—at the high, middle, or lower end? If priced at the high end, how will the pricing be positioned in relation to the primary competitors? What positioning message will the pricing communicate?

Pricing at the low end of the market (in competition with Ward Systems, SPSS, etc.) will require enormous volume to generate a significant revenue stream. While this is possible, especially from a global perspective, it will also require enormous marketing resources to generate awareness and interest.

In addition, the perceived value of a product priced at the low end will likely be less in the marketplace, regardless of its performance. For example, it may be quite possible that the predictive modeling capability of the products from Ward Systems Group may be superior to that which is offered by SAS. However, there is also a perception that SAS is better because of its cost and the “cache” associated with its price.

A new company built around a superior predictive model can price itself higher than comparable competition, about the same, or lower. Each strategy has its strengths and weaknesses. If the new company prices higher, it must position itself as being worth the additional cost because of perceived or actual benefits. If the product is truly demonstrably superior, then the pricing can be justified and will likely hold up in the marketplace.

But the key to a higher pricing strategy will depend on the value the customer places on those benefits that are highly motivating. In other words, if a clearly superior model is important to a prospective customer because of the money it will save, the money it will make or both, then higher pricing will hold up. If the perceived value of the superior benefit is not as highly motivating, then the pricing will not hold up and the new company may be at a competitive disadvantage.

Pricing comparable to the competition can be advantage when the prospective customers are highly price sensitive, thus keeping price out of the picture and allowing any decision to made on the merits of the benefits. At the same time, a competitor’s pricing strategy may be flawed, either because of inadequate cost considerations or because of a misread of the marketplace. A lower price, on the other hand, can be a competitive advantage when the products are comparable and the lower price can be justified because of a lower cost structure. In fact, a lower price and a superior product could be just the right dynamic combination to signal superior value to the prospective customer. However, lower price means lower margins and can be especially damaging if a higher priced could be gained from a customer.

24

Page 25: Bearse Final

A new predictive model company should strive to position itself as worth more to the prospective customer. Therefore, a higher price can be justified and successful.

4. Penetration A fourth decision relates to penetration. What will be the primary penetration strategy? Will the product/service be sold using a direct selling model or are there superior alternative strategies that are more efficient and cost-effective?

A direct sales penetration strategy appears to be the most costly, yet fruitful, strategy for a new predictive model product. While the Internet provides tremendous opportunities to distribute product, including software, a direct selling approach is likely to be the best strategy for determining the needs and values of prospective customers and securing commitment. In addition, it is also highly likely that the software will need to be customized for each customer, thus requiring significantly more individualized attention. The marketplace is also used to and expects personal attention at the high-end software-selling environment.

Costs are higher, however, in a direct selling environment. In addition to sales commissions and/or salaries, other expenses include travel, training, recruitment and back-office support. A new company would also have to likely have to secure the services of an experienced sales staff, which would put salaries at a minimum of $150K per year and higher.

If the new model is priced at the low-end, then the Internet would be the ideal distribution method. Not only would costs be lower, but users would likely have plenty of experience in this system and be comfortable with its use.

5. Positioning A fifth marketing decision relates to positioning. How will the product be positioned? What succinct message will be communicated to prospective customers that indicates the product is different and better than the competition? What key unique selling proposition, benefit, feature, or attitude will be communicated?

Aside from determining the optimal target market, positioning is the single most important marketing component. Get the positioning and target markets right and the rest will ultimately follow. By positioning, we mean the succinct message planted in the minds of the customer that says how the product is different and better than the competition. Perhaps more importantly, the key to successful positioning is finding a dimension that is highly motivating to the target market that you can offer which the competition does not and cannot offer.

At first glance, just having a superior predictive model should be the basis for any positioning strategy. However, a better product doesn’t always translate into a superior position. For example, Microsoft didn’t necessarily have a superior operating system in the early 80’s. But because it’s positioned itself as the

25

Page 26: Bearse Final

“standard” operating system for business and because standardization was highly motivating to IT professionals in the business environment, Microsoft capitalized on the opportunity.

Quantitative marketing research can be useful and helpful in determining the optimal positioning strategy for a new predictive modeling company. While expensive when performed correctly (100-150K), marketing research can save a new company time and money by avoiding false paths, expensive trial-and-error, and wasted marketing communication strategies, all of which may doom a startup before every getting “out of the box.”

In summary, it is impossible to say what the positioning strategy should be for a new predictive modeling company without a better understanding of the needs, problems, and motivations of prospective customers, coupled with a full exploration of possible communication strategies. While a superior predictive model may indeed be the defining attribute that carries a positioning strategy, it may also be only incidental to other more critical elements.

6. Beta Testing A sixth marketing decision relates to beta testing. To what extent will the new model be tested to what length? In other words, what level of confirmed superiority will need to be demonstrated in order to successfully launch the product?

A beta test is clearly in the best interest of a new predictive model. The efficacy of the model, especially when demonstrated in “real world” situations, will enhance the selling and marketing environment. In addition, several different scenarios are recommended where the new model is pitted against the main or top competitors for each industry. Thus, if SAS is considered the heavyweight in the financial industry, then the beta test should be able compare results of SAS and the new model in similar real world environments. A note of caution—these results may or may not form the basis for an effective positioning strategy. However, they can be considered extremely useful, helpful and, therefore, necessary.

7. Marketing ResearchA seventh marketing decision relates to marketing research. What level of qualitative and quantitative research will be required to adequately and accurately identify the optimal positioning, targeting, pricing and communication strategies prior to and after launch?

Quantitative marketing research is strongly advised to pinpoint target market and positioning opportunities. While secondary research can be useful in providing an overview of the potential markets, only primary research can provide the level of detail and sophistication that is required to accurately choose target market and positioning strategies.

26

Page 27: Bearse Final

Risk FactorsThere are several risk factors that could affect the outcome of a successful launch of a predictive modeling company. Those factors are:

What if a continued decelerating economy (slipping into recession) tightens up financial resources of prospective customers?

What if a larger competitor, such as SAS, comes out with a comparable product and/or service?

What if current customers of data mining and predictive modeling products remain satisfied with their current vendor and don’t feel the need to switch?

What if a beta test does not show a statistically significant improvement between the new predictive model and current models?

What if professional jealousy and envy within the data mining and predictive modeling industries (including academia) thwarts attempts to gain wide acceptance of new model?

What if in-house statistical efforts by most, if not all, of the database management, data warehousing, data mining, and statistical software companies prevents opportunities to license new model?

What if competitive pressures continue to force prices downward, thereby reducing profit margins?

What if tight venture capital and other fundraising markets thwart efforts at much needed capital to develop and grow new business?

What if concerns over privacy, especially on the Web, force data mining efforts to be curtailed?

What if overcrowded market conditions prevent a new entry from gaining quick and extensive brand awareness without having to apply significant marketing resources?

What if an inadequate sales staff, either in terms of training or geographic coverage, limits the sales penetration strategy?

What if an undifferentiated positioning strategy forces a new entrant to compete solely on price?

What if a poor pricing strategy could not only reduce sales, but lower margins? What if choosing the wrong optimal target market slows sales, adds costs, and

forces a new company to redirect its marketing efforts, thereby losing precious time and momentum in the marketplace?

Further Work and Research NeededWhile it seems apparent that there may be a significant market for a new predictive modeling company, further work and research is still needed. The following is a summation of questions still needing answers.

To what degree are the current customers of data mining companies, such as SAS, satisfied with the products and services being offered?

What combination of motivating messages and/or benefits would be necessary to drive them away from their current provider and towards a startup?

27

Page 28: Bearse Final

Which of the available markets (financial, telecom, etc.) offer the greatest revenue and profit potential, especially in the beginning? Which offer the least potential?

Which of the available markets offer the best opportunity to enter with the least competition?

Who precisely are the “buyers” of predictive modeling? Is the ideal target market in the marketing or IS departments, or is the better candidate the chief executive officer of a company?

What size of company can be considered an ideal candidate for predictive modeling services?

Are there any markets not listed in this report that offer significant opportunity? What pricing strategies for the respective markets chosen offer the greatest

opportunity for revenue enhancement and profitability? What value by prospective customers on a new predictive model can be ascertained?

How will the positive results of beta testing affect the marketing of a new predictive model?

What impact will a prolonged decelerating economy have on the market potential of a new predictive model?

What if a strong competitor (such as SAS) came out with a comparable predictive model? How would that affect the sales, marketing and business strategy of a new company?

Which companies currently operating in the data mining, data warehousing, and predictive modeling environments are the optimal candidates for possible partnerships or suitors for a new company? At what price should a possible sale be worth considering?

Which is the better strategy—licensing the predictive model for use at the “wholesale” level or direct distribution at the “retail” level?

Which competitors pose the greatest threat to a new predictive modeling company and in which markets?

28

Page 29: Bearse Final

Conclusion

Every business starts with a need or problem—a need or problem that a customer has to solve or satisfy. The bigger the need or problem, the better the business opportunity. One of the continuing needs faced by many companies in this day and age of increasing data accumulation is to make sense out of that data, and then use that knowledge to make better business decisions.

Because of this ongoing and critical need, data mining has emerged as an important business function. Thus, it shouldn’t be surprising that companies are now spending enormous amounts of dollars in this arena and will continue to do so in the near future.

At the same time, companies continue to desire to improve their data mining and predictive capabilities. Thus, it would stand to reason that any new company that could improve (perhaps dramatically so) the predictive abilities of its data mining function would be in high demand. In fact, the evidence is clear that there is significant opportunity. One need only look at the success of SAS and 2-year-old NuTech Solutions as evidence of the enormous potential resting in this marketplace.

But that also means the competition remains fierce and, in some cases, bloody. This is not a market for the faint-of-heart. Large, well-established firms and young upstarts alike continue to carve up an attractive pie, leaving little for the companies that fail to match the needs of the marketplace with well-positioned products or services.

A new entrant can compete successfully. But it must identify optimal targets (ideally that aren’t being well served by the competition), correctly position itself as being different and better in dimensions that are highly motivating to its target markets, and it must successfully execute an effective marketing communications strategy, whether that is through a direct sales force, advertising, promotions, public relations or some combination.

In order to accurately determine where the marketing opportunities lie, a new entrant into this market would be wise to use quantitative marketing research first before launching into the marketplace. While expensive when done correctly, this will result in far better pinpointed efforts immediately, thereby preventing costly trial and error that may in result in partial success or outright failure—both of which can be fatal to startups.

In addition, it is highly suggested that a systematic beta test occur, which would be designed to test and prove the efficacy of the new predictive model, especially in comparison to competitive offerings. While successful beta test results alone are not likely to be the basis for a comprehensive marketing strategy, they do provide a level of credibility that can form the basis of a foundation from which to operate. “Show me” is likely going to be the common refrain from prospective customers, especially if the new entrant declares superiority against well-established competitors in the marketplace.

29

Page 30: Bearse Final

These beta tests must demonstrate not only superior effectiveness, but must also be attached to a return-on-investment that prospective customers can fully appreciate.

*********************In summary, despite possibly entering a highly competitive field, a business built around a new (and presumably superior) predictive model holds tremendous promise and potential. More examination must be done to accurately determine the best markets and distribution channel for such a business. In addition, quantitative marketing research may be necessary to pinpoint the business opportunities. However, with the continued interest in data mining by companies of all shapes and sizes, and with the potential high profit margins associated with this kind of intellectual property, a successful predictive modeling business can potentially generate significant revenue and profit in a relatively short period of time, assuming many of the inherent risks are mitigated.

30